Real-Time Recognition of Dielectric Elliptical Object Parameters Using an ANN Method Based on EM Pulse Scattering

This paper introduces a novel artificial neural network (ANN) model designed to reconstruct the shape, permittivity, and rotation angle of dielectric elliptical objects. The model uses the Method of Auxiliary Sources (MAS) to solve scattering problem efficiently, eliminating field singularities and speeding up the solution process. Gaussian periodic pulse feedback is used to provide data for the ANN’s estimations. The ANN can make real-time recognitions, with a typical $5 \%$ variation in intermediate parameters and a $0.1 \%$ average deviation on training data. Since it only requires a single measurement, the model is suitable for real-time applications like subsurface imaging and missile tracking. This work establishes a new framework for integrating machine learning with computational electromagnetics, enabling advancements in defence systems and imaging technologies.